Dissociation of cyanogen azide: An alternative route to synthesis of carbon nitride

Article abstract of DOI:10.1021/jp9818479

Solid films composed principally of carbon and nitrogen were grown on a variety of substrates at ambient temperature in a flow-tube reactor by upstream mixing of cyanogen azide, cyanogen, or cyanogen halides with active nitrogen obtained from an electrical discharge. Ab initio calculations and dependence of deposition rates on both choice of donor and N atom production suggests that NCN radicals are a critical growth species. The films obtained are electrically insulating with a refractive index of 2.3 at visible wavelength and are optically transparent from 550 nm out to at least 14 ??m with the exception of two broad absorption bands centered at 1550 and 3250 cm^-1, the latter band growing in upon exposure of the film to atmospheric moisture. Film analysis by X-ray photoelectron spectroscopy revealed comparable concentrations of both carbon-to-nitrogen bonds (with approximate C₃N₄ stoichiometry) and diamond-like carbon-to-carbon bonds as well as minority bonding of carbon to impurities.

Full text of DOI:10.1021/jp9818479

6
010
J. Phys. Chem. B 1998, 102, 6010-6019
Dissociation of Cyanogen Azide: An Alternative Route to Synthesis of Carbon Nitride
D. J. Benard,* C. Linnen, and Alan Harker
Rockwell Science Center, Thousand Oaks, California 91358
H. H. Michels and J. B. Addison
United Technologies Research Center, East Hartford, Connecticut 06108
R. Ondercin
Air Force Wright Materials Laboratory, Wright-Patterson AFB, Ohio 45433
ReceiVed: April 14, 1998; In Final Form: June 4, 1998
Solid films composed principally of carbon and nitrogen were grown on a variety of substrates at ambient
temperature in a flow-tube reactor by upstream mixing of cyanogen azide, cyanogen, or cyanogen halides
with active nitrogen obtained from an electrical discharge. Ab initio calculations and dependence of deposition
rates on both choice of donor and N atom production suggests that NCN radicals are a critical growth species.
The films obtained are electrically insulating with a refractive index of 2.3 at visible wavelength and are
optically transparent from 550 nm out to at least 14 µm with the exception of two broad absorption bands
-1
centered at 1550 and 3250 cm , the latter band growing in upon exposure of the film to atmospheric moisture.
Film analysis by X-ray photoelectron spectroscopy revealed comparable concentrations of both carbon-to-
nitrogen bonds (with approximate C
3
N
4
stoichiometry) and diamond-like carbon-to-carbon bonds as well as
minority bonding of carbon to impurities.
Introduction
Over the past decade, our laboratory has investigated dis-
sociative reactions of the halogen azides (XN3, X ) F or Cl),
which are useful for powering short-wavelength chemical
1
Cohen intially predicted that carbon nitride (C3N4) could exist
in a stable hexagonal phase analogous to â-Si3N4 with hardness
greater than natural diamond, and many investigators have
explored a variety of physical methods to prepare thin films of
this interesting new material. Typically, small carbon clusters
and energetic nitrogen ions are obtained from separate sources
and co-deposited onto a common substrate in a suitable vacuum
chamber. Deposition onto heated substrates has also been used
to inhibit formation of softer thermally unstable paracyanogen
structures as well as to promote crystallinity in the films obtained
by activating the surface layer and destabilizing more weakly
6
,7
lasers.
Collisional transfer of modest amounts of thermal,
vibrational, or electronic energy to these parent or donor
molecules results in prompt fragmentation and efficient libera-
tion of metastable electronically excited NX* radicals. Also,
once a critical threshold is achieved, the dissociation process
becomes autocatalytic due to feedback of released energy from
excited products to residual azides by an energy-transfer
mechanism. In the case of FN3 dissociation, feedback of
released vibrational energy (V) occurs from primary byproducts
(FN3 f N2(V) + NF*) as well as secondary products of
2
-5
bound amorphous growth sites at the reactive interface.
collisional quenching (M + NF* f M(V) + NF) and self-
annihilation reactions (2NF* f N2(V) + 2F) of the metastable
Thermal control, however, cannot be applied to all substrate
materials of interest. Finally, since every C atom in C3N4 is
bonded exclusively to four adjacent N atoms, physical methods
are characterized by endothermic insertion of N atoms into C-C
bonds that either are present in the original clusters or form
upon growth of the film.
8
9,10
species. Coombe and Ray have studied analogous dissocia-
tion mechanisms in ClN3 which liberate electronically excited
metastable (NCl*) radicals, and these authors provide a similar
explanation for the chained decomposition.
On the basis of commonalities in the above research data, it
seemed feasible that a rich source of NCN could also be obtained
from analogous dissociation of the (X ) CN) pseudohalogen
azide. Prior matrix isolation and photolysis studies of cyanogen
azide (NCN3) by Milligan and later by Krogh also suggested
Because of difficulties inherent to physical N atom insertion,
we have sought a chemical synthesis in which the critical C-N
bonds are established prior to film deposition. From this point
of view, a precursor should contain only N and C atoms with
no direct C-C bonds and full coordination of all carbon bonds
to nitrogen, as well as an atomic N/C ratio greater than 4/3 and
thermodynamic potential greater than C3N4. The NCN radical
meets these criteria and is thought to be a superior growth
species, since formation of the nitride from this intermediate is
thermodynamically driven by elimination of gaseous nitrogen
11,12
this possibility.
Rapid self-annihilation of NCN radicals via
the reaction (2NCN f 2CN + N2), however, constitutes a
potential difficulty in this scheme that is also implied by analogy
to the known kinetics of NF* and NCl* decay following
6
,7
dissociation of the corresponding parent azides.
(
N2) which escapes from the growing film while steric factors
In this work, a primary stream of helium (He) diluent and
N2 passes through an electrical discharge before mixing on-
simultaneously hinder creation of new C-C bonds.
S1089-5647(98)01847-1 CCC: $15.00 © 1998 American Chemical Society
Published on Web 07/15/1998

Dissociation of Cyanogen Azide
J. Phys. Chem. B, Vol. 102, No. 31, 1998 6011
the-fly with a secondary stream of donor NCN3 molecules and
He diluent in a low-pressure flow-tube reactor. Electronically
TABLE 1: Calculated Electronic States of NCN @ R(C-N)
)
1.2 Å
a
b
or vibrationally excited metastable (N2*) molecules and N atoms
symmetry
E (hartrees)
T (eV) /T (eV)
e
e
produced in the discharge1
3-17
react with the injected donor and
3
-
X Σ
g
-146.708 863 9
-146.685 798 1
-146.684 729 9
-146.588 427 2
-146.585 059 2
-146.583 390 0
-146.549 873 8
-146.529 292 2
-146.517 589 5
-146.502 692 8
0.00/0.00
0.63/1.42
0.66/1.45
3.28/4.07
3.37/3.37
3.41/3.41
4.33/4.33
4.89/5.68
5.21/5.21
5.61/6.40
liberate dissociation products that grow solid films of high
carbon and nitrogen content on a substrate suspended down-
stream in the flow. Alternate donors including cyanogen
1
a ∆
g
1
+
b Σ
g
1
-
c Σ
u
3
+
A Σ
u
(
(
NCCN), cyanogen chloride (ClCN), and cyanogen bromide
BrCN) were investigated in like manner. Issues germane to
3
B ∆
u
3
C Π
u
this coating process include the relative importance of N atoms
versus N2* molecules in reaction with the donor species, the
roles of NCN vs CN as species potentially related to film growth
at the substrate, and characterization of the resulting solid film
product. Similar carbon-nitrogen films have only recently been
deposited in our laboratory at substrate temperatures up to 375
1
d Π
u
3
D Π
g
1
e Π
g
a
CASSCF (10, 6)/6-311+G(d) level of theory. b Singlets shifted
1
by +0.79 eV to match B3LYP/6-311+G(2df) results at the a ∆ state.
g
°
C. For the most part, the films described herein were formed
Electronic States of NCN. Excluding core electrons, the
at ambient temperature and are therefore not optimized with
respect to crystallinity, adhesion, or mechanical characteristics
molecular orbital configuration of NCN with the lowest
4
2
2
2
3
electronic energy is (1πu) (4σg) (3σu) (1πg) [X Σg] and low-
lying excited states are formed by variations of the 1πg, 3σu,
and 1πu occupancy. To achieve balance in calculation of the
(hardness and surface friction) that will be addressed in
subsequent publications.
ground and low-lying excited states of NCN, a full CASSCF
Theoretical Section
32,33
wave function
was constructed for the 1πu, 3σu, 1πg, and
Covalent azides, of the XN3 type discussed above, have
4σg orbitals. The distribution of 10 valence electrons among
six valence orbitals resulted in 15 triplet-state configurations
and 21 singlets. Several basis sets were explored, and the Pople
relatively weak central bonds6
,7,18-22
and dissociate to form
electronic ground-state N2 and corresponding halogen or
pseudohalogen nitrene (NX) fragments along the lowest avail-
able potential energy surface. Since these azides and N2 both
have singlet ground states, while halogen nitrenes have triplet
ground states, XN3 dissociates adiabatically into the first
34
6-311+G(d) triple-ú + polarization + diffuse basis was chosen
as a balance between complexity and computer time.
Both the ground and low-lying excited states of NCN are
3
5
known to be linear and symmetric. Preliminary calculations
demonstrated the NCN ground and excited states to share a
common potential minimum, and a fixed N-C bond length of
1.2 Å was chosen. The calculated spectra, given in Table 1,
1
electronically excited NX(a ∆) state which is metastable, while
ground-state NX radicals correlate with excited states of the
3
azide via a repulsive A′′ potential energy surface. The excited
1
triplet states of N2 lie too high relative to the azide ground state
show the lowest singlet state to have ∆g symmetry analogous
2
3,24
3
-
3
to play any significant role in the dissociation process.
to NCl and NF. The (000-000) bands for the X Σg -C Πu
1
1
A weak potential barrier on the singlet surface separates the
azide ground state from its correlated NX* + N2 products, and
adiabaticity of the dissociation reaction depends critically on
an intersection of this surface with the repulsive triplet state. If
the crossing occurs inside the potential well, as in HN3, then it
and a ∆g-d Πu transitions of NCN are found experimentally
3
5,36
at 3.77 and 3.73 eV, respectively.
Our calculated term
values for these transitions are about 12% too large, as the
CASSCF procedure does not properly account for correlation
effects, which contribute most significantly to errors in the
singlet-triplet energy difference. The results in Table 1 should
therefore be viewed as a general guide to locating electronically
excited states of NCN, and more detailed calculations using
correlated wave functions are required for a precise determi-
is encountered many times due to molecular vibrations before
surmounting the barrier to dissociation.2
5,26
These conditions
result in a loss of adiabaticity and formation of ground-state
triplet NX products via a low probability per encounter tunneling
mechanism. When the singlet-triplet crossing occurs on the
product side of the barrier, it is encountered only once as the
molecule dissociates, and consequently metastable singlet NX*
products are formed with near unit quantum efficiency.
3
-
1
nation. Since the intersystem term value for the X Σg -a ∆g
transition is not known experimentally, calculations for these
two symmetries were also carried out at a higher level of theory.
37
Clifford et al. recently reported ab initio results of calcula-
6
,7
-
Previous theoretical and experimental investigations have
demonstrated that adiabatic dissociations of FN3 and ClN3 occur
over reaction barriers of approximately 0.5 and 0.7 eV,
respectively. In the case of FN3, the critical singlet-triplet
crossing occurs on the product side of the reaction barrier,
tions on NCN and NCN using correlated wave functions for
3
-
1
both the X Σg and a ∆g states of NCN at three different levels
3
8,39
of theory based on the CBS approximations.
In the order
of increasing accuracy, these authors found T00 ) 1.525, 1.286,
and 1.249 eV for the CBS-4, CBS-Q, and CBS-QCI/APNO
methods, respectively. The latter method is among the most
accurate currently available with typical rms errors as low as
0.8 kcal/mol. Our calculations of the lowest triplet and singlet
states of NCN were carried out at the B3LYP/6-311+G(2df)
1
resulting in exclusive production of NF(a ∆), whereas in ClN3
the crossing occurs nearer to the peak of this singlet potential
energy surface, and both excited singlet and ground-state triplet
27
NCl dissociation products are obtained. Although there have
been previous theoretical studies2
9-31
of NCN3, the nature of
level of theory. The density functional method of Becke
40
potential energy surfaces leading to dissociation products has
not been investigated in detail, and there is a lack of prior gas-
phase experimental results bearing on this process. Conse-
quently, to aid the present investigation, two of the authors
(B3LYP), with an extended 6-311+G(2df) basis set, was also
37
-
used successfully in previous studies of NCN and NCN . We
37
used the basis set previously employed by Clifford et al. in
-
-
their analysis of the NCN and HNCN ions. At this level of
3
-
1
(Michels and Addison) have carried out an ab initio study of
theory, T00 ) 1.42 eV was obtained for the X Σg -a ∆g term
values, indicating an accuracy close to the CBS-Q method. As
a first approximation, the last column in Table 1 lists uncorre-
lated term energies that have been corrected upward by a
the lowest singlet and triplet surfaces of NCN3 as well as a
preliminary investigation of resulting NCN excited states that
are formed by the dissociation reaction.

6012 J. Phys. Chem. B, Vol. 102, No. 31, 1998
Benard et al.
constant 0.79 eV (for singlets only) to match our correlated result
1
at the a ∆g energy level.
Thermochemistry. The heat of formation of cyanogen azide
can be estimated from a calculated heat of the decomposition
1
3
-
1
+
reaction (NCN3[ A′] f NCN[X Σg ] + N2[X Σg ]) since heats
of formation for both the NCN radical and N2 are known
41
experimentally. The corresponding isogyric reaction (NCN3-
1
1
1
+
[
A′] f NCN[a ∆g] + N2[X Σg ]) is less certain since, as
discussed above, the NCN singlet-triplet term value has not
been established experimentally. The first dissociation reaction
3
-
(
leading to NCN[X Σg ] product) was therefore examined at
both the B3LYP/6-311+G(2df) and CBS-Q levels of theory,
the latter being accurate to ∼1.3 kcal/mol for molecules of
similar complexity. Our calculated enthalpy change for the
decomposition reaction is -11.3 kcal/mol (25 °C), based on
CBS-Q calculated energies. Combining this result with a known
heat of formation of +108.2 kcal/mol for the NCN radical, we
find for NCN3 that ∆Hf(25 °C) ) +119.5 kcal/mol. Our
calculated enthalpy based on B3LYP/6-311+G(2df) energies,
∆
Hf(NCN3, 25 °C) ) +119.6 kcal/mol, is also in close
agreement with the CBS-Q prediction. These results are typical
of the heats of formation previously calculated for FN3 and
7
ClN3 and are in reasonable agreement with spectroscopic
estimates by Okabe and Mele.22
Barrier Dynamics. Activation energy for decomposition of
NCN3 was calculated at both the CBS-Q and B3LYP levels of
theory. Correcting for zero-point energy, we find transition-
state barriers of 1.19 and 1.18 eV (relative to the minimum in
3
Figure 1. Singlet and triplet potential surfaces of NCN relevant to
molecular dissociation.
the active species were determined from pressure measurements
and known vapor pressure versus temperature curves42 of the
bromide. Two methods were employed to generate batches of
NCN3 similarly diluted in He buffer gas using the same stainless
steel cylinder as a reaction vessel. The first method involved
adding ClCN gas to a suspension of solid sodium azide (NaN3)
1
the A′ surface for NCN3) at the CBS-Q and B3LYP levels of
theory, respectively. These calculated barrier heights are
significantly larger than previously determined values for FN3
and ClN3, in qualitative agreement with experimental evi-
6
,7,12
dence
that NCN3 has greater thermal stability (decay time
on the order of a week at ambient temperature relative to hours
or fractions thereof for ClN3 and FN3, respectively). On the
basis of the calculated barrier height, it appears that NCN3 can
be efficiently dissociated in single collisions with vibrationally
excited N2 molecules that are either at or above the V ) 4 energy
4
3
in an anhydrous aprotic polar solvent. Propylene carbonate
C4H6O3) was the solvent of choice due to its desirable chemical
(
4
2
properties and high (240 °C) boiling point which facilitated
both the generating reaction as well as disengagement of the
gaseous azide product from solution without significant hydro-
carbon contamination. In the second method, NCN3 was
generated by direct reaction of BrCN with NaN3 in the absence
of a solvent, upon heating a mixture of the solids slightly above
2
3
level.
The potential energy surface of ground-state NCN3 is
illustrated in Figure 1, based on optimization of all bond lengths
1
and bond angles, at fixed NCN-N2 separation in A′ molecular
4
2
5
2 °C to melt the bromide.
symmetry, at the B3LYP/6-311+G(2df) level of theory. The
These reagents, and the products generated from them, are
3
repulsive A′′ surface was subsequently calculated using
4
4
highly toxic. Appropriate care must therefore be taken to
avoid ingestion, inhalation, or skin exposure. Additionally,
grinding of NaN3 crystals may initiate rapid decomposition45
into reactive sodium metal and N2 gas, while upon condensation
from the gas-phase NCN3 forms a highly unstable explosive
parameters optimized for the singlet state. Figure 1 illustrates
1
3
the A′ and A′′ surfaces of NCN3 crossing near the singlet
surface maximum of the dissociation barrier, similar to ClN3.
1
We therefore expect metastable singlet NCN(a ∆g) molecules
to be the most favored reaction product, but some nonadiabatic
4
3
liquid (vapor pressure ∼ 70 Torr @ 300 K). Cyanogen is
3
-
branching to ground-state NCN(X Σg ) is possible.
also a high-energy fuel capable of intense combustion with air
1
Production of NCN(a ∆g) by dissociation of NCN3 may have
44
or other oxidizing species. The reagents cyanogen, cyanogen
chloride, and cyanogen bromide are also highly corrosive.
Tubing, valves, fittings, reactors, gauges, and storage tanks
should all be constructed of stainless steel, glass, or inert
a profound effect on the growth of C3N4 films, if these
metastable species propagate in substantial yield to the substrate.
The additional electronic excitation carried by such precursors
can aid passage over opposing reaction barriers at the gas-
solid interface, and the paired electron spin symmetry of this
precursor may open other more favorable reactive channels as
well.
4
4
polymers such as Teflon. Finally, use of an oversized (300
cfm) vacuum pump with suitable discharge treatment and
frequent oil changes is also warranted to prevent exposure to
toxic exhaust fumes.
Approximately 12.5 g of NaN3 was added to 120 mL of
solvent in the reactor, which also contained a 2 in. long Teflon-
coated magnetic stirring bar. The stainless steel cylinder, with
an absolute pressure gauge attached, was then quickly pumped
down to vacuum and back-filled with ClCN gas (to 1 atm
pressure) with the magnetic stirrer turned off. The cylinder was
Experimental Section
Preparation. Mixtures of NCCN, BrCN, or ClCN in He
diluent gas were prepared in a stainless steel cylinder (2.5 in.
i.d. × 18 in. long) using neat reagents supplied by commercial
manufacturers without further purification. Mole fractions of

Dissociation of Cyanogen Azide
J. Phys. Chem. B, Vol. 102, No. 31, 1998 6013
Figure 3. Schematic of the flow tube reactor used for deposition of
carbon-nitrogen films.
Figure 2. IR absorption spectrum of process gas stream (chloride-
based NCN generator).
3
replaced by comparable BrCN peaks.47 In either case, removal
of bypassed halogen cyanides from the process gas stream by
then sealed off and stirring reactivated to absorb typically 90%
of the chloride gas into the solvent in approximately 1-2 min.
Subsequently, stirring was interrupted and the gas charging
process repeated three more times, before finally adding He
diluent to a net pressure of 80 psig. Stirring was then resumed
for 2-3 h, and the lower third of the sealed-off cylinder warmed
to 70 °C by means of external heating tapes. The azide diluted
in He was finally withdrawn from the reservoir (still at 70 °C,
with continued stirring) via a corrosion-resistant gas regulator
adjusted to dispense at atmospheric pressure. This gas flow
was in turn directed to a 14 cm long stainless steel FT-IR
absorption cell with salt windows that evacuated to a down-
stream needle valve adjusted to bleed the cylinder pressure at
use of an in-line cold trap was not practical due to lower vapor
42,43
pressure of the azide,
and use of excess NaN3 in the reactor
(
to consume the cyanide) unfortunately diminished the yield of
12
NCN3. From this last observation and the appearant stability
of NCN3 in contact with dry NaN3, one may conclude that NCN3
has at least a deleterious slow reaction with NaN3 that is
promoted when the solid azide is dissolved either in the solvent
or BrCN melt.
Flow Tube. Figure 3 is a schematic illustration of the
experimental apparatus used to grow carbon-nitrogen films at
ambient temperature. The flow tube was constructed of 1.375
in. i.d. quartz tubing and was operated at pressures in the range
2
5 psig/h. The first hour of gas removal from the reaction
1
0-15 Torr. A converging/diverging nozzle made of glass-
cylinder was discarded to vacuum to ensure complete saturation
of conductive tubing, after which the flow of dilute NCN3 and
He was redirected into the flow tube. The bromide method
was carried out in the same reaction vessel using 1 g of BrCN
and 2 g of NaN3 with 80 psig of added He per batch, and
reaction was complete after 2-3 h at 60 °C.
filled Teflon (area ratio ∼ 8) was positioned typically midway
between the injector probe tip and a coaxially mounted 1 in.
diameter substrate. The nozzle served to concentrate reactive
gas flow along the centerline, thereby increasing interaction with
the substrate and effectively eliminating wall losses. Assuming
parabolic flow, an approximate centerline velocity of 65 m/s
(upstream of the nozzle) was estimated from the cross sectional
area, measured gas pressure, and a pressure versus flow curve
of the vacuum system that was obtained separately by use of a
calibrated electronic mass flowmeter. Pressure and velocity
downstream of the nozzle depended on the position of the
movable azide injector probe in relation to the nozzle throat.
With the probe tip located upstream of the nozzle, upstream
and downstream pressures varied less than 10%, indicating
subsonic flow through the nozzle and centerline gas velocities
downstream of the nozzle comparable to upstream values.
Moving the 0.375 in. o.d. injector tube into or through the nozzle
throat significantly restricted its open area and lead to pressure
ratios as high as 3 to 1 (upstream to downstream) with
corresponding acceleration of the gas to supersonic velocity at
Figure 2 presents a typical FT-IR absorption spectrum of the
process gas stream from the chloride-based generator. With
-
1
the exception of two unresolved peaks at 444/450 cm (not
shown in these data) all of the anticipated NCN3 bands, based
on original matrix isolation work11 by Milligan, are present at
2
249/2208, 2163/2151, 2101, 1263/1247, 923, 865, and 655
-
1
46
cm . Principal impurity peaks derived from carbon dioxide
-1
(CO2) at 3650, 2350, and 630 cm , the latter band overlapping
4
7
NCN3, and residual ClCN, which also has two strong bands
-1
obscuring weaker azide peaks at 2249/2208 and 865 cm . Since
the main absorption band of the C4H6O3 solvent46 at 1862 cm
was not detected, and there is little evidence of peaks due to
hydrocarbons in the vicinity of 3000 cm , thermal separation
of azide gas from the propylene carbonate solvent was highly
-1
46
-1
-
1
48
effective. The narrow 1263/1247 cm bands of NCN3 are
easily isolated and serve as a convenient diagnostic for monitor-
ing the yield of gaseous azide. Typically, this absorption rises
as total reservoir pressure diminishes during the course of an
experiment, indicating a relatively constant partial pressure of
azide gas inside the reservoir, resulting from gas equilibration
with NCN3 still dissolved in the propylene carbonate solution.
The bromide-based generator was more convenient to use
and more stable in terms of azide delivery during the course of
an experiment but produced somewhat lower yields of NCN3.
The CO2 impurity previously caused by solvent decomposition
was also effectively eliminated in the bromide generator which
results in a cleaner FT-IR spectrum with ClCN impurity peaks
the nozzle exit plane. Partial pressure of the donor species
(NCN , NCCN, ClCN, or BrCN) in the flow tube was
3
approximately 1 mTorr (upstream) based on total pressure
measurements and mass flow rates. Rapid injection of donor
molecules into the primary flow was optimized by using a
conical probe tip (30° half-angle) with 12 0.035 in. diameter
radial holes drilled around its circumference (normal to the
reactor centerline) and by adding a variable flow of He to the
secondary gas stream to allow penetration of resulting mixing
jets into the surrounding stream of N2*/He. Visually observing
spatial dependence of resulting chemiluminescence demonstrated
that mixing of primary and secondary flows was complete within
1 cm of the injector tip.

6
014 J. Phys. Chem. B, Vol. 102, No. 31, 1998
Benard et al.
Discharge. Two methods were employed for production of
interference rings (rainbows) that appeared from the edge to
the center. A more precise interferometric method involves
measuring VRS data at near-normal incidence in the center of
the film. Smooth oscillations in the VRS signal appear as a
result of Etalon effect in the film with a peak-to-peak amplitude
ratio dependent on film index and with periodicity dependent
on both index as well as thickness according to known principles
the active nitrogen used to dissociate donor molecules. In the
first method, a prepared mixture of 3-5% N2 in He buffer gas
was admitted through a 1/4 in. i.d. × 9 in. long quartz side
tube, where a continuous electrical discharge was energized by
a constant-current (6 kV/20 mA/60 Hz) neon sign transformer
between concentric electrodes fashioned from viton O-ring
sealed stainless steel tube unions. Prior to addition of azide,
the above discharge filled the flow tube with an orange afterglow
that is characteristic of N atom recombination.17 Previously
generated thin carbon-nitrogen films could be removed from
their substrates by overnight exposure to this active nitrogen
gas stream, and all substrates were cleaned in situ for several
hours by this method before deposition of new films.
5
1
of optics which have been employed in commercial instru-
52
ments. A two-parameter fit of the VRS data to optical theory
therefore determines both index and depth of the film. In
practice, it is necessary to use a somewhat more complicated
model based on the same concept, which also takes into account
the substrate index and reflection from its back surface to obtain
the most accurate results. Some thickness results were nonethe-
less verified, adding certainty to our index determination, upon
use of surface profilometry to measure the step height of films
deposited over a removable tape mask. Comparative ITS
measurements were also used to estimate depth of heavier films
based on calibration of the absorption data to prior thin film
results.
Substituting a more easily ionized buffer gas such as argon
Ar) for half of the He in the side tube discharge reduces both
electron temperature and the yield of N atoms (in relation to
(
N2*) since the excited molecules have a lower excitation
potential than dissociated atoms.17 Changing buffer gas in this
manner also significantly attenuates the characteristic orange
chemiluminescence; however, complete substitution of Ar for
He leads to discharge filamentation and less efficient volumetric
interaction with the gas stream in the side tube. The three-
component gas mixture was therefore used as one method to
favor production of excited N2* molecules over N atoms.
Results
Deposition. All tests were done with substrates mounted
normal to the flow and coaxially centered in the reactor to
produce near circular concentric interference rings in the finished
films, as mounting of substrates inclined to the flow resulted in
complicated interference patterns that were difficult to analyze
for thickness. Semitransparent carbon-nitrogen films were
readily grown on 1 in. diameter stainless steel, quartz, sapphire,
diamond, silicon (100), silicon carbide, gallium nitride, gallium
arsenide, zinc sulfide, and Cleartran substrates. Adhesion on
polished silicon was poor, possibly due to surface oxidation,
and these films rapidly peeled away from the substrate upon
exposure to atmosphere. Attempts to remove various films on
the other substrates by means of abrasion with a pink rubber
eraser yielded mixed results; however, some films were obtained
on stainless steel, zinc sulfide, and silicon carbide that were
strongly adherent. Deposition onto silicon carbide at 230 °C
improved both adhesion and stability of the film against peeling;
however, this trend was found to reverse at still higher substrate
temperatures. Significant differences in film depth were also
obtained using a series of identical zinc sulfide substrates while
holding the rate of donor injection and run time approximately
constant upon varying donor type, discharge configuration,
injector probe position, mixing rate, and substrate location in
the flow tube.
An alternate discharge generator was also employed using
identical He-based gas flows as described above by wrapping
a 10 cm long copper foil around the outside of the flow tube in
a region upstream of the Teflon nozzle. A power amplifier
applied a 50 kHz/1 kV (rms) sine wave drive to this capacitively
coupled external electrode in relation to the grounded stainless
steel injector tube which passed through it, and at this power
level the resulting radial discharge in the flow tube produced
only a faint violet-blue afterglow from the N2/He gas stream,
indicating a significantly reduced yield of N atoms. This
afterglow emission, caused by N2(A) self-annihilation, was
short-lived and died out over a distance of only a few
centimeters in the flow.1
7,49
The discharge conditions employed in either of the generators
described above are comparable to those of continuous wave
carbon dioxide lasers that operate in He-N2-CO2 gas mixtures
at similar levels of total pressure and He dilution by efficient
5
0
excitation of N2 ground-state vibrational energy levels. By
comparison, the corresponding excitation rates of both atoms
and electronically excited states under these conditions is
negligible. Measurements of both current and voltage indicated
similar electrical power inputs to the two discharge generators.
Production of thick films was strongly favored by each of
the following factors: azide donor, deposition in subsonic flow,
Diagnostics. Visible and near-ultraviolet chemiluminescence
spectra (VCS) were taken of the activated gas stream in a fixed
region between the nozzle and substrate as a function of donor
addition and movable injector location using a scanning (1200
line/mm) grating spectrometer at 5 nm resolution with a 1P28
or GaAs photomultiplier tube, picoammeter, and chart recorder
for detection. Upon completion of a typical 2 h coating run,
substrates were removed for study by visible reflectance/
transmission spectroscopy (VRS/VTS), infrared transmission
spectroscopy (ITS), X-ray diffraction/photoelectron spectroscopy
mixing optimized for complete utilization of available N *, and
2
low N atom (radial) discharge configuration. Films estimated
at 4-8 µm thick were produced in nominal 2 h coating runs
following the above prescription. Adjusting the He flow that
was added to the donor gas stream through the mixing nozzle
also had significant beneficial impact on film uniformity over
the substrate surface as well as subsequent stability against
peeling upon atmospheric exposure. Films obtained from ClCN
were only marginally detectable, whereas comparable films
grown from NCCN or BrCN were each more readily observed
but still 5-10 times thinner than films grown from equimolar
quantities of azide donor.
(
XRD/XPS), and scanning electron microscopy (SEM) to
determine optical, chemical, and crystallographic properties.
Electrical conductivity of the films was measured by a four-
point probe, and several related methods were employed to
determine film thickness. For relatively thin films, which only
partially covered the substrate and were dome-shaped, thickness
could be estimated by counting the number of visible wavelength
In contrast to the above observations, production of films from
NCCN donor optimized using the side tube discharge with pure
He diluent and film depth was significantly attenuated (in this
case) if He was partially substituted by Ar to attenuate N atoms.

Dissociation of Cyanogen Azide
J. Phys. Chem. B, Vol. 102, No. 31, 1998 6015
Figure 4. Visible emission spectrum of N
by side tube discharge.
2
(B-A) afterglow produced
Figure 5. Visible emission spectrum of N (B-A) afterglow produced
2
by radial discharge.
This apparant change of direction versus the azide donor
described above can be explained if NCN radicals are required
for film growth as follows. Since N2(V) carries too little energy
to dissociate NCCN and is present in much larger concentration
than N atoms, the fast gas-phase reactions most likely to follow
injection of azide donor with a weak N atom source (N2(V) +
NCN3 f 2N2 + NCN) and addition of cyanogen donor with a
more intense N atom source (N + NCCN f NCN + CN) both
result in production of NCN. Use of cyanogen donor in the
presence of copious N atoms, however, is complicated by
deleterious side reactions such as (N + NCN f N2 + CN) and
length in the flow tube even with suppression of the N atom
afterglow, while substituting BrCN, ClCN, or NCN3 in place
of cyanogen produced pale orange, green, and yellow chemi-
luminescent emissions, respectively, that were each long-lived
in the flow tube. Recordings of VCS in the absence of donor
injection demonstrated characteristic emissions of the N2(B-
A) first positive band system in the >470 nm wavelength
region, with a great majority of the chemiluminescence
intensity beyond visual range in the near-infrared. Donor
addition yielded no new emission peaks that were not also found
in the discharged N2 afterglow without donor present, with the
2
3
(N + CN f N2 + C), the latter of which is known to occur at
sole exception of isolated narrow tail bands at 385 nm assigned
53
55
gas kinetic rate. Consequently, potential contributions of both
C atoms and CN radicals to films grown from cyanogen cannot
be completely eliminated.
to the CN(B-X) transition.
Consequently, the visually
observable effects of various donors on the chemiluminescence
spectrum were all caused by changing intensity distributions
within the vibrational manifolds of excited N2(B) molecules.
The CN(B-X) intensity initially scaled linearly with addition
of NCCN donor and eventually saturated due to depletion of
N2*. A similar scaling relationship was obtained with use of
the azide donor until the concentration of NCN3 (as measured
Spectroscopy. Partial pressure of N2 in these experiments
(
∼1 Torr) is high enough to ensure vibrational equilibration of
the metastable A state with its ground electronic state. Con-
sequently, changes in the vibrational distribution of ground-
state N2 are at least qualitatively reflected in the B-A
chemiluminescence which derives from metastable self-an-
-
1
by absorption at 1263/1247 cm ) rose above a critical level in
the region of ∼10% peak attenuation, beyond which the intensity
of CN(B-X) chemiluminescence increased markedly. As
mentioned in the Introduction, threshold behavior for the onset
of autocatalytic decomposition is anticipated from this energetic
donor due to chain reactions that feed back released chemical
energy to residual azide molecules.
4
9
nihilation. Figure 4 presents an assignment of vibrational
2
3
transitions emitting within the N2(B-A) band system using
emission data taken in the vicinity of the substrate with no
donors present and side tube discharge excitation of the gas
stream. In this case, an approximately quasi-thermal vibrational
distribution is obtained with little relative population above V
5
6
)
6. The vibrational temperature, however, is still large
No measurable NCN(A-X) emission at 329 nm was
detected, and narrow bands of NCN analogous to the well-
known vertical a-X and b-X forbidden intercombination
compared to ambient temperature of the gas stream, illustrating
negligible loss due to slow V-T relaxation. Use of the radial
discharge in place of the side tube generator produced the data
that are similarly analyzed in Figure 5. Here, the vibrational
distribution is significantly nonthermal, with suppression of the
yield at V ) 3 and excess relative population at both V ) 7 and
V ) 11 in evidence. Undoubtedly, strong pumping of the N2
6,7,23
transitions
of NF or NCl were not observed within the range
of detector sensitivity for any of the donors studied. On the
basis of correlated ab initio results presented earlier, the a-X
and b-X transitions of NCN are expected in the spectral vicinity
of 900 nm; however, these bands are also anticipated to be very
weak due to higher molecular symmetry of NCN, which adds
a parity selection rule to the spin and orbital angular momentum
constraints that also forbid such NX transitions in general. Since
these very weak NCN transitions unfortunately fall into the same
spectral region as the intense and ubiquitous N2* discharge
afterglow, their detection is most problematical.
5
0
4
vibrational energy levels occurs in either discharge generator
and initially produces a significant nonthermal population
5
which equilibrates in transit due to rapid V-V energy transfer.
Consequently, the closer coupling and superior yield of vibra-
tionally excited N2 molecules (at or above the V ) 4 energy
level) associated with use of the radial discharge accounts in
large part for the higher deposition rates obtained in this
discharge configuration with use of the azide donor.
In contrast to the above discussion, the vicinity of allowed
NCN(A-X) emission is spectroscopically clear, and intuitively
one might expect that if nearly resonant CN(B-X) emission
was excited, for example, by energy transfer from metastable
Reaction between discharged N2 and NCCN donor produced
essentially no observable chemiluminescence at visible wave-

6016 J. Phys. Chem. B, Vol. 102, No. 31, 1998
Benard et al.
N2(A), then some 329 nm chemiluminescence should also be
2
3
in evidence. Transferring 6.2 eV of electronic energy from
N2(A) results in dissociation of NCN as opposed to CN
chemiluminescence, however, because the reactions (NCN f
N2 + C and NCN f CN + N) have respective thermodynamic
4
0
energy thresholds of 2.5 and 4.3 eV while at least 7.4 eV are
required to dissociate CN. Similar considerations also apply
to the known parallel N2(A) + NCN3 f 2N2 + NCN(A)
5
7
reaction and the as likely parallel N2(A) + RCN f R + N2
CN(B) processes in which R ) Cl, Br, or CN. These
+
reactions are each less exothermic than the corresponding direct
energy transfers described above; however, as shown in the
Theoretical Section the heats of formation of NCN3 and NCN-
(X) only differ by about 0.5 eV, which is insufficient to stabilize
the former reaction against branching into additional dissociative
product channels. The branching ratio from N2(A) into NCN-
(A) is therefore expected to be very small compared to similar
excitations of CN(B) in either case. Consequently, the presence
or lack of NCN radicals in the flow tube cannot be reliably
evaluated from observations of spontaneous chemiluminescence,
and more definitive methods such as laser-induced fluorescence
are required instead to make this determination. Finally, while
N2(A) may play a role in exciting some observable short
wavelength emission, it has negligible influence on film growth.
Figure 6. Visible transmission spectrum of C/N film deposited on
Cleartran substrate.
Once the dissociation reaction is complete, collisional energy
transfer processes are expected to support a state of local
thermodynamic equilibrium in which all excited species decay
at a common rate on longer time scales. The N2* decay
following injection of NCCN was monitored at fixed residence
time in the flow (T ∼ 1.5 ms) by detecting the relative intensities
of CN(B-X) emission as a function of initial donor concentra-
tion [D]. Fitting this emission data to the form [CN(B)]/[D] )
A exp - (K[D]T), where A is an arbitrary constant, yields an
Due to limited production in the discharge and very fast self-
annhilation49 N2(A) is limited to concentrations that are small
compared to the added donor.
Kinetics. Based on selection rules already discussed, radia-
1
tive decay of metastable NCN(a ∆) is expected to be as slow
-
12
3
effective N2* quench rate of K ∼ 6 × 10
cm /(molecule s).
or slower than the analogous 0.2/s58 and 0.7/s
7,59,60
transitions
Based on this result and typical flow tube conditions (T ) 1
in NF and NCl, respectively. Spontaneous emission therefore
contributes negligible loss of NCN* while in transit to the
substrate. Consequently, the balance of metastable versus
ground-state NCN radicals arriving at the film surface depends
solely on collisional relaxation and reactive phenomena.
1
3
3
ms and [D] ) 3 × 10 /cm for coating at 2-4 µm/h), the
product K[D]T representing fractional loss of N2* currently has
a value of ∼20%. Lower rates of film growth associated with
use of NCCN donor in the present reactor setup could therefore
be partially compensated by a 5-fold increase of the donor mole
fraction before completely depleting the N2* bath. Equivalent
concentration scaling of NCN3, however, should lead to growth
rates as high as 10-20 µm/h or higher, as in this regime
autocatalytic dissociation of the energetic donor enhances N2*
faster than it is depleted.
The observed strong preference for azide donor, as a source
of NCN in the absence of significant N atoms, requires sufficient
kinetic lifetime in active flow for these radicals to propagate to
the substrate before substantial removal by self-annhilation. This
condition, in turn, sets an approximate upper limit on the
magnitude of a corresponding NCN self-annihilation rate
constant (k). If t is taken as transit time in active flow between
the injector probe and substrate, then the fraction of NCN
molecules lost to self-annihilation depends on a dimensionless
product term, k[NCN]t. Since this term cannot significantly
exceed unity, consistent with the mechanism and observations
discussed above, one then finds with t ∼ 1 ms and [NCN] ∼
Optical. On transparent substrates, the deposited films had
a faint yellow-brown coloration. As shown in Figure 6, VTS
data revealed a soft band edge that is characteristic of thin films
with an onset of visible wavelength absorption at 550 nm,
corresponding to an approximately 2.6 eV optical band gap.
The refractive index at visible wavelength and the depth of a
relatively thin film deposited onto a zinc sulfide substrate were
determined from VRS data shown in Figure 7 to be 2.3 and 0.8
µm, respectively, by fitting the 7° off-normal reflectance
spectrum to a computer model based on the theory of dielectric
layers, which also included effects of backside reflection from
the substrate. The computer model was exercised using both
index and thickness of the film as independent fit parameters
to match period, phase, and amplitude of spectral oscillations
in the reflectance data caused by film interferences. Secondary
solutions were also obtained using an additional third fit
parameter which uniformly scaled the known substrate index
across the visible (400-800 nm) wavelength range. A satisfac-
tory (unique) solution was indicated when both the three-
parameter fit optimized with a substrate index scaling factor
close to unity and similar values were obtained for the film
index and thickness as in the original two-parameter solution.
Subsequent profilometry measurements, given as Figure 8, also
yielded a film depth of ∼0.8 µm, in good agreement with our
1
3
3
-11
3
[
initial donor] ∼ 3 × 10 /cm that k < 3 × 10 cm /(molecule
s). For comparison, the rates of metastable self-annihilation in
NF and NCl have been previously measured as approximately
-
12
-11
3
6,7,61
3
× 10
and 3 × 10
cm /(molecule s), respectively.
At low levels of donor injection, the initial rise of CN(B-
X) chemiluminescence tracks the rate of parent molecule
dissociation. The rise of CN(B) was spatially resolved by
varying injector probe position in relation to a fixed observation
zone, and these data were found to approximately correlate with
corresponding variations of film thickness. Dissociation time
constants for trace concentrations of NCN3 and NCCN were
measured as approximately 0.2 and 0.5 ms, respectively. Failure
to grow films in supersonic flow may have also resulted due to
formation of a bow shock48 over the substrate; however, based
on the above dissociation rates, transit times were too small in
any case to permit efficient use of either donor under these
conditions.

Dissociation of Cyanogen Azide
J. Phys. Chem. B, Vol. 102, No. 31, 1998 6017
Figure 7. Visible reflectance spectrum of C/N film on ZnS substrate
Figure 10. Low-resolution XPS scan of film grown from azide donor
and process impurities.
(×××) fit to computer model (s) used for determination of refractive
index and film depth.
atmosphere until placed in the spectrometer, after which repeated
scans demonstrated that progressive reaction of the film with
atmospheric moisture contributed to growth of the shorter
wavelenth hydride stretch band. The longer wavelength bands,
assignable to C-C and C-N stretching, are similar to those
seen in carbon nitride films produced by co-deposition from
6
4
separate nitrogen and carbon sources.
Apart from differences in overall intensity of the bands due
to variances in film depth, comparable ITS signatures were also
obtained using samples produced from cyanogen that were
relatively free of significant impurities, as the only other
materials used in depositing these thinner films were He and
N2. Therefore, incorporation of impurities related to NCN3 such
as chlorine, bromine, or oxygen do not appear to significantly
alter IR absorptions present in the heavier films made using
azide chemistry. The extent to which such impurities may have
contributed to reaction with atmospheric moisture, and indirectly
to either film peeling or growth of absorptions caused by
formation of air reaction products, still remains to be determined.
Chemical. A low-resolution XPS scan of a film produced
from azide donor, obtained via the chloride-based generator, is
given as Figure 10. These data show the film to be composed
principally of carbon and nitrogen, but also significant amounts
of impurity oxygen (15%) and chlorine (7%) as well as trace
concentrations (<1%) of zinc and sodium. Undoubtedly,
incorporation of substantially large amounts of chlorine into the
film derives principally from the presence of unreacted ClCN
in the flow tube as monitored by the FT-IR diagnostic. The
oxygen impurity may derive either from a similar CO2 impurity
that is related to solvent breakdown in the (chloride-based) azide
generator or from surface reaction of the film upon exposure to
atmospheric moisture.
Figure 8. Film depth by surface step profilometry (C/N film deposited
over a tape mask).
The overall N/C ratio of 0.53 obtained from XPS data
indicates the presence of excess carbon that is either bonded to
other carbon or impurity atoms in the film. Higher resolution
XPS data show all the N atoms to be in equivalent sites, but
the carbon line is split and deconvolves, as shown in Figure
11, to a triad of components that can be assigned to C atoms
that are bonded to either other C atoms, N atoms, or impurities
such as O or Cl atoms. The small central peak is easily
identified with impurity sites, while comparison of the peak
centered at 285 eV to similar studies of diamond-like carbon
shows this signal to be caused by C-C bonding.65 The
remaining carbon peak at 288 eV is therefore assignable to C-N
Figure 9. IR transmission spectrum of ∼8 µm thick C/N film
(corrected for substrate).
interferometric result. The measured 2.3 refractive index of
62,63
these films is comparable to high-grade diamond-like carbon.
Infrared transmission spectroscopy of the films demonstrated
two broad and featureless absorption bands centered at 1550
-
1
and 3250 cm , as shown in Figure 9. The ITS data presented
here were generated from a thick film deposited onto zinc
sulfide, and the measurement was referenced to and corrected
for transmission characteristics of an identical bare substrate.
The newly deposited film was maintained in a dry nitrogen
6
6
bonds. Area analysis of the deconvoluted peaks shows that

6018 J. Phys. Chem. B, Vol. 102, No. 31, 1998
Benard et al.
Figure 11. Deconvolution of 288 eV XPS data into single lines
corresponding to C atoms that are bonded to other C atoms, N atoms
or impurities.
Figure 12. Scanning electron micrograph of stress-cracked C/N film
on diamond substrate.
roughly 40% of the C atoms are bonded to N atoms and the
N/C ratio for this carbon component is approximately 1.3,
relatively close to the anticipated (4/3) stoichiometry. Conse-
quently, the films can be approximately characterized as near-
equal mixtures of diamond-like carbon and stoichiometric C3N4
with lesser impurities bonded to carbon depending on the
specific method of preparation. A similar analysis of oxygen-
free films obtained from azide donor via the bromide generator
demonstrated a single N atom line as before, no change in C-C
bond density, and a corresponding increase of C-N bonding,
suggesting that incorporated oxygen substitutes for nitrogen in
these films.
Because individual film plates do not lie flat on the substrate,
it is not possible at present to perform reliable nanoindentation
measurements for quantitative determination of the bulk modu-
lus. More work is therefore required to improve film adhesion
to the substrate and eliminate stress cracking before such
definitive hardness measurements can be attempted. Improve-
ments in crystallinity are also indicated in order to achieve full
theoretical hardness.
Conclusions
Physical. The films obtained were electrically insulating
The N/C ratio of films produced from NCN3 is comparable
(>10 MΩ/square @ 0.7 µm thickness). No observable discrete
6
4
to other methods previously reported.
Advantages of the
peaks, other than from the substrate, were seen in XRD
measurements of films produced at ambient temperature. Long-
range crystallographic order is therefore not present in these
cases, and the films are best characterized as amorphous.
Examination of the films by SEM, however, demonstrated
microscopic cracking along straight lines with cracks generally
meeting at right angles and individual plates curled up away
from the substrate, as shown in Figure 12. These qualitative
observations respectively suggest films composed of a relatively
hard material with latent (short-range) crystal order that is
initially formed in tension on the substrate.
method are relatively high growth rates in simple apparatus and
ease of scalability to large volume production. Evidence of
NCN as a key intermediate in the process remains inferential,
and laser-induced flourescence or sensitive absorption measure-
ments of this and other related species are required to tie down
a detailed understanding of the gas-phase kinetics. Based on
results of ab initio calculations, a substantial yield of singlet
metastable NCN radicals is obtained from dissociation of the
azide donor. The azide dissociation reaction is driven primarily
by collisional energy transfer from vibrationally excited N2
molecules at or above the V ) 4 energy level, and it is
anticipated that a significant fraction of the resulting electroni-
cally excited NCN radicals reach the substrate where differences
in spin symmetry or internal energy can play a significant role
with respect to film growth. The relative importance of ground
versus excited states of NCN as potential growth species,
however, still needs to be evaluated experimentally. Results
obtained to date also illustrate a highly reactive growth surface
capable of incorporating impurities related to the presence of
More recent X-ray measurements performed on films grown
on either silicon carbide or gallium nitride at deposition
temperatures near 375 °C have each revealed sharp diffraction
peaks at lattice spacings of 1.498 and 2.703 Å that are not
present from the associated bare substrates. These data were
obtained with good signal-to-noise ratio and precisely cor-
respond to the two strongest [301] and [200] reflections of
6
7
â-phase C3N4 previously observed by Johnson.
Purely
crystalline films have not yet been established by intensity of
the peaks observed, and it is possible that isolated microscopic
crystalline domains (in an otherwise amorphous matrix) account
for these results. Correlated with this finding, however, we also
found that growth of the short wavelength IR absorption band
due to hydride formation was also significantly attenuated in
films deposited at higher substrate temperature, suggesting that
a more densified film structure was obtained in these cases.
Details regarding these most recent findings will be included
in subsequent publications.
bypassed ClCN and possibly byproduct CO in the flow tube.
Consequently, to the extent that CN radicals are produced in
2
the reactor by NCN self-annihilation, excess carbon is likely to
accumulate in the resulting film because the unity N/C ratio of
this growth species is below the preferred C N stoichiometry,
3
4
and this condition necessarily results in formation of nonvolatile
byproducts. Simultaneous optimization of both film growth rate
and N/C ratio therefore requires setting the residence time in
the active flow as well as donor concentration consistent with
the dissociation time constant and kinetic rate coefficient of
64

Dissociation of Cyanogen Azide
J. Phys. Chem. B, Vol. 102, No. 31, 1998 6019
NCN self-annihilation. Surface preparation and substrate tem-
perature during deposition were not treated as significant
variables in this initial study, and work is in progress to optimize
film adhesion as well as crystallinity. Finally, related physical
properties such as surface friction and hardness will in time
require a detailed mechanical characterization.
(27) Benard, D. J.; Winker, B. K.; Chowdhury, M. A.; Seder, T. A.;
Michels, H. H. Proc. Int. Conf. Lasers-1991; STS Press: 1992.
(28) Murgich, J.; Aray, Y. J. Chem. Phys. 1987, 87, 3580.
(29) Tsuda, M.; Oikawa, S.; Nagayama, K. Photogr. Sci. Eng. 1983,
27, 118.
(30) Eisenstein, M.; Hirshfeld, F. L. Chem. Phys. 1979, 38, 1.
(
(
(
(
31) Bak, B.; Jansen, P.; Stafast, H. Chem. Phys. Lett. 1975, 35, 247.
32) Hegarty, D.; Robb, M. A. Mol. Phys. 1979, 38, 1795.
33) Schlegel, H. B.; Robb, M. A. Chem. Phys. Lett. 1982, 93, 43.
34) Krishnan, R.; Binkley, J. S.; Seeger, R.; Pople, J. A. J. Chem. Phys.
Acknowledgment. The authors are indebted to their col-
leagues Les Warren, Randy Hall, Bill Southwell, Ellen Boehmer,
Dave Marshall, Dave Deitrich, and Gary Pollack, who provided
many useful discussions on synthetic chemistry as well as
assistance with a variety of film diagnostic methods. This work
was supported by contracts with Anteon Corporation, the Air
Force Wright Laboratory Materials Directorate, and the Air
Force Office of Scientific Research.
1980, 72, 650.
(35) Herzberg, G.; Travis, D. N. Can. J. Phys. 1964, 42, 1658.
(
36) Kroto, H. W.; Morgan, T. F.; Sheena, H. H. Trans. Faraday Soc.
970, 66, 2237.
37) Clifford, E. P.; Wenthold, P. G.; Lineberger, W. C.; Petersson, G.
1
(
A.; Ellison, G. B. J. Phys. Chem. 1997, 101, 4338.
(38) Montgomery, J. A.; Ochterski, J. W.; Petersson, G. A. J. Chem.
Phys. 1994, 101, 5900.
(39) Ochterski, J. W.; Petersson, G. A.; Montgomery, J. A. J. Chem.
Phys. 1996, 104, 2598.
References and Notes
(40) Becke, A. D. J. Chem. Phys. 1993, 98, 5648.
(
41) Chase, M. W.; et al. JANAF Thermochemical Tables, 3rd ed.; J.
1) Cohen, M. L. Phys. ReV. 1985, B12, 7988.
Phys. Chem. Ref. Data 1985, 14, Suppl. 1.
42) Weast, R. C. Handbook of Physics and Chemistry, 60th ed.; CRC
Press: Boca Raton, FL.
(
3
(
(
43) Marsh, F. D.; Hermes, M. E. J. Am. Chem. Soc. 1964, 86, 4506.
44) Lewis, R. J. Dangerous Properties of Industrial Materials, 9th ed.;
1
Van Nostrand: New York.
(45) Material Safety Data Sheet (Sodium Azide, Sigma Chemical).
(46) Pouchert, C. J. Aldrich Library of FT-IR Spectra; Aldrich Chemi-
cal: Milwaukee, WI, Vol. 3.
(47) Freitag, W. O.; Nixon, E. R. J. Chem. Phys. 1965, 24, 109.
(48) Liepman, H. W.; Roshko, A. Elements of Gas Dynamics; Wiley:
New York, 1957.
(
8) Benard, D. J. J. Appl. Phys. 1993, 74, 2900.
(49) Clark, W. G.; Setser, D. W. J. Phys. Chem. 1980, 84, 2225.
(50) Nighan, W. L.; Bennett, J. H. Appl. Phys. Lett. 1969, 14, 240.
(51) Born, M.; Wolf, E. Principles of Optics, 5th ed.; Pergammon: New
York.
(52) Filmetrics, Model F20 (www.filmetrics.com).
(53) Whyte, A. R.; Phillips, L. F. Chem. Phys. Lett. 1983, 98, 590.
(9) Ray, A. J.; Coombe, R. D. J. Phys. Chem. 1994, 98, 8940.
(
(
10) Ray, A. J.; Coombe, R. D. J. Phys. Chem. 1995, 99, 7849.
11) Milligan, D. E.; Jacox, M. E.; Bass, A. M. J. Chem. Phys. 1965,
3, 3149.
12) Krogh, O. D.; Ward, C. H.; Hollenbeck, J. M. J. Phys. Chem. 1982,
6, 2892.
13) Stedman, D. H.; Setser, D. W. J. Chem. Phys. 1970, 52, 3597.
14) Setser, D. W.; Stedman, D. H.; Coxon, J. A.; J. Phys. Chem. 1970,
3, 1004.
15) Balamuta, J.; Golde, M. F. J. Chem. Phys. 1982, 76, 2431.
16) Golde, M. F.; Moyle, A. M. Chem. Phys. Lett. 1985, 117, 375.
17) Boehmer, E.; Ph.D. Dissertation, Max Planck Institute Fur Stro-
mungsforschung, Gottingen, 1990, ISSN 0436-1199.
18) Patel, D.; Pritt, A. T.; Benard, D. J. J. Phys. Chem. 1986, 90, 1931.
19) Christen, D.; Mack, H. G.; Schatte, G.; Willner, H. J. Am. Chem.
Soc. 1988, 110, 707.
20) Coombe, R. D.; Patel, D.; Pritt, A. T.; Wodarczyk, F. J. J. Chem.
Phys. 1981, 75, 2177.
4
8
(
(
54) Trainor, C. E.; Rich, J. W.; Rehm, R. G. J. Chem. Phys. 1968, 48,
(
(
1
798.
(55) Tatematsu, S.; Kuchitsu, K. Bull. Chem. Soc. Jpn. 1977, 50, 2896.
5
(
9.
56) Beaton, S. A.; Ito, Y.; Brown, J. M. J. Mol. Spectrosc. 1996, 178,
(
(
(
9
(
57) Meyer, J. A.; Klosterboer, D. H.; Setser, D. W. J. Chem. Phys.
1
971, 55, 2084.
(
(
(
(
58) Malins, R. J.; Setser, D. W. J. Phys. Chem. 1981, 85, 1342.
59) Becker, A. C.; Schurath, U. Chem. Phys. Lett. 1989, 160, 586.
60) Yarkony, D. R. J. Chem. Phys. 1987, 86, 1642.
(
(
61) Quinones, E.; Habdas, J.; Setser, D. W. J. Phys. Chem. 1987, 91,
(
5
155.
(
(
62) Moravec, T. J.; Orec, T. W. J. Vac. Sci. Technol. 1981, 18, 226.
63) Bendow, B., Ed. Proceedings of the DARPA Workshop on
(
(
(
21) Kroto, H. W. J. Chem. Phys. 1966, 44, 831.
22) Okabe, H.; Mele, A. J. Chem. Phys. 1969, 51, 2100.
23) Rosen, B. Selected Constants RelatiVe to Diatomic Molecules;
Diamond-Like Carbon Coatings; Albuquerque, April 1992; Contract MDA
903-81-C-0151.
Pergammon; New York, 1970.
(
24) Herzberg, G. Electronic Spectra of Polyatomic Molecules; Van
Nostrand: New York, 1966.
25) Schmidt, M. W.; Gordon, M. S.; Dupuis, M. J. Am. Chem. Soc.
985, 107, 2585.
26) Kajimoto, O.; Yamamoto, T.; Fueno, T. J. Phys. Chem. 1979, 83,
29.
(64) Hu, J.; Yang, P.; Lieber, C. M. Phys. ReV. 1998, B57, R3185.
(65) Belton, D. N.; Harris, S. J.; Schmeig, S. J.; Weiner, A. M.; Perry,
T. A. Appl. Phys. Lett. 1989, 54, 416.
(66) He, Z.; Carter, G.; Colligon, J. S. Thin Solid Films 1996, 283, 90.
(67) Johnson, D. J.; Chen, Y.; He, Y.; Prince, R. H. Diamond Relat.
Mater. 1997, 6, 1799.
(
1
(
4